Abstract
In order to improve robustness of remote sensing algorithms for lakes, it is vital to understand the variability of inherent optical properties (IOPs) and their mass-specific representations (SIOPs). In this study, absorption coefficients for particulate and dissolved constituents were measured at 38 stations dis-tributed over a biogeochemical gradient in Lake Balaton, Hungary. There was a large range of phytoplank-ton absorption (aph(k)) over blue and red wavelengths (aph(440) 5 0.11–4.39 m21, aph(675) 5 0.048–2.52 m21), while there was less variability in chlorophyll-specific phytoplankton absorption (a*
ph(k)) in the
lake (a*ph(440) 5 0.022 6 0.0046 m2mg21, a*ph(675) 5 0.010 6 0.0020 m2mg21) and adjoining wetland
sys-tem, Kis-Balaton (a*ph(440) 5 0.017 6 0.0015 m 2
mg21, a*ph(675) 5 0.0088 6 0.0017 m 2
mg21). However, in the UV, a*ph(350) significantly increased with increasing distance from the main inflow (Zala River). This was
likely due to variable production of photoprotective pigments (e.g., MAAs) in response to the decreasing gradient of colored dissolved organic matter (CDOM). The slope of CDOM absorption (SCDOM) also increased
from west to east due to larger terrestrial CDOM input in the western basins. Absorption by nonalgal par-ticles (aNAP(k)) was highly influenced by inorganic particulates, as a result of the largely mineral sediments
in Balaton. The relative contributions to the absorption budget varied more widely than oceans with a greater contribution from NAP (up to 30%), and wind speed affected the proportion attributed to NAP, phy-toplankton, or CDOM. Ultimately, these data provide knowledge of the heterogeneity of (S)IOPs in Lake Balaton, suggesting the full range of variability must be considered for future improvement of analytical algorithms for constituent retrieval in inland waters.
1. Introduction
Understanding the sources and magnitude of variability in light absorption in lakes and reservoirs is funda-mentally important to studies concerned with photochemistry [Moran and Zepp, 1997; Bertilsson and Tranvik, 2000], photosynthesis [Blache et al., 2011], primary production [Tilstone et al., 2005; Lee et al., 2011], heat and energy transfers [Jolliff et al., 2008; Dera and Wozniak, 2010], and biogeochemical models [Ciavatta et al., 2014]. The absorption and scattering of light (termed inherent optical properties, IOPs) are also key processes influencing the magnitude and spectral distribution of the water-leaving reflectance signal meas-ured by Earth-observing satellites [Mobley, 1994; Kirk, 1994]. Remote sensing has allowed for characteriza-tion of water bodies at improved spatial and temporal scales with the aim of monitoring water quality operationally in ocean, coastal and, more recently, inland waters. However, remote sensing algorithms developed for retrieval of physical and biogeochemical properties in open ocean waters are often inaccu-rate when applied to more turbid and optically complex inland waters [Sathyendranath et al., 1999; IOCCG, 2000; Binding et al., 2008]. Inland waters typically have higher concentrations of phytoplankton biomass, detritus, inorganic particulates and color dissolved organic matter (CDOM), and large percentages of sus-pended particulates can be land derived. Moreover, the biogeochemical properties of inland waters do not covary over space and time, resulting in potentially large variability in the IOPs of the optically active con-stituents (OACs) [Binding et al., 2008; Palmer et al., 2015]. In order to improve the performance of algorithms for the retrieval of biogeochemical parameters in lakes, it is vital that we develop a better understanding of the variability in the absorption and backscattering coefficients (a(k) and bb(k); m
21
) of the main OACs in
Correspondence to: C. A. L. Riddick, [email protected] Citation: Riddick, C. A. L., P. D. Hunter, A. N. Tyler, V. Martinez-Vicente, H. Horvath, A. W. Kovacs, L. V€or€os, T. Preston, and M. Presing (2015), Spatial variability of absorption coefficients over a biogeochemical gradient in a large and optically complex shallow lake, J. Geophys. Res. Oceans, 120, 7040–7066, doi:10.1002/ 2015JC011202.
Received 5 AUG 2015 Accepted 6 OCT 2015
Accepted article online 8 OCT 2015 Published online 31 OCT 2015
VC2015. American Geophysical Union. All Rights Reserved.
lakes (i.e., phytoplankton, nonalgal particles (NAP), and colored dissolved organic matter (CDOM)) and their mass-specific representations (a*(k) and bb*(k); m2mg21). When light enters the water column photons are
removed from its path by absorption, and a(k) is defined as the sum of absorption by particulate and dis-solved constituents and water itself. Light is also scattered by suspended particles, and the scattering coeffi-cient (b(k)) is commonly defined as a measure of the total magnitude of scattered light (without regard to its angular distribution). bb(k) can therefore be defined as the total light scattered in the backward direction.
The sum of a(k) and b(k) is the beam attenuation coefficient (c(k)), or the total light attenuated in the water column. Knowledge of the IOPs is particularly important for radiative transfer studies and the development of analytically based inversion algorithms. The bio-optical properties of open ocean [Morel and Maritorena, 2001] and coastal waters [Babin et al., 2003b] have been extensively studied over the last four decades, but our knowledge of these properties in lakes and other inland waters remains comparatively poor [Luis Perez et al., 2011; Zhang et al., 2011], particularly for highly turbid and productive water bodies.
In spite of the fact that lakes represent only a small fraction of the Earth’s total surface water (0.013%) [Shiklomanov, 1993], the variability in their absorption and scattering coefficients is likely to be far greater than that encountered in the oceans, shelf seas, and coastal waters because the close proximity of land. Sur-face runoff from land exerts a strong influence on the composition and concentration of dissolved and par-ticular matter in lakes. In turn, absorption and scattering by dissolved and particulate materials affect the spectral shape and magnitude of the remote sensing reflectance (Rrs(k)) measured by satellite sensors
[Morel and Prieur, 1977; Kirk, 1994]. Knowledge of the variability of mass-specific inherent optical properties (SIOPs) is thus necessary for the interpretation of water-leaving reflectance signals. More formally, Rrscan be
related to a(k) and bb(k) via equation (1), where f is an experimental constant dependent on the light field
and volume scattering function and Q is a parameter accounting for geometrical attenuation of light exiting the water column [Gordon et al., 1975; Morel and Gentili, 1991; Dall’Olmo and Gitelson, 2005]:
Rrsð01; kÞ / f Q bbð Þk a kð Þ1bbð Þk (1)
The absorption and backscattering coefficients can be further partitioned into the contributions from each optically active constituent via equations (2) and (3),
a kð Þ5aphð Þ1ak NAPð Þ1ak CDOMð Þ1ak wð Þk (2)
bbð Þ5bk b;phð Þ1bk b;NAPð Þ1bk b;wð Þk (3)
where aph(k), aNAP(k), aCDOM(k), and aw(k) represent the absorption coefficients for phytoplankton, NAP,
CDOM, and water, and bb,ph(k), bb,NAP(k), and bb,w(k) represent the backscattering coefficients for
phyto-plankton, NAP, and pure water, respectively. It is generally assumed that CDOM is nonscattering.
The relative contribution of the dissolved and particulate constituents to absorption and backscattering budgets varies significantly between different water types. Moreover, changes in size and composition of the constituents can result in marked changes in the SIOPs. For instance, the mass-specific particulate scat-tering coefficient [b*p(555)] was reported to vary in ocean and coastal waters based on the proportion of
organic versus mineral particles, and particle water content, apparent density, and refractive index [Babin et al., 2003a]. The variability of SIOPs is a major source of uncertainty in the interpretation of water-leaving reflectance signals, in the retrieval of biogeochemical properties and in estimates of primary production [Dall’Olmo and Gitelson, 2005; Gilerson et al., 2010; Tilstone et al., 2012]. The improved parameterization of remote sensing algorithms for turbid lakes and other optically complex waters thus relies on a comprehen-sive knowledge of the mass-specific absorption and scattering coefficients of lake water constituents and on an understanding of the sources and magnitude of their variability across different lake types. In particu-lar, knowledge of the absorption properties at the wavelengths of440, 620, and 675 nm have direct impli-cations for the remote sensing retrievals of phytoplankton pigments, including chlorophyll-a (Chl-a) and phycocyanin (PC), in order to distinguish potentially harmful cyanobacteria blooms in lakes [Simis et al., 2005; Kutser et al., 2006; Hunter et al., 2010; Mishra et al., 2013].
The (S)IOPs of inland waters have been relatively poorly studied, but the limited research to date suggests that they exhibit significant variability. Luis Perez et al. [2011] found that the absorption and mass-specific absorption coefficients of particulate matter in Laguna Chascomus, Argentina show significant seasonal
blooms [Matthews and Bernard, 2013]. Spatial differences in optical properties within a single water body have also been reported, although chiefly with regard to IOPs rather than SIOPs. Following a wind event in western Lake Erie, spectral variations in a(k) and its contributing components was reported, and more mod-est wavelength dependencies for bp(k) and bbp(k) which were consistent with observations reported for
coastal systems [O’Donnell et al., 2010]. Similarly, spatial heterogeneity in IOPs and apparent optical proper-ties (AOPs) has been reported in Lake Champlain, with uncoupled variation between absorption and bio-geochemical parameters [O’Donnell et al., 2013]. A study by Effler et al. [2012] demonstrated both spatial and temporal differences in the IOPs in Oneida Lake, NY, with the sum of aNAP, aCDOM, and aphat 440 nm
ranging from 0.9 to 2.0 m21over the summer months (June–August). This study further found a high
contri-bution from CDOM to a(440), however variations in NAP and phytoplankton ultimately drove absorption dynamics. Thus, while it has been widely documented in previous studies that IOPs and AOPs are variable in both time and space along with variations in the optically active constituents (OACs), further knowledge is required to characterize the relationships between IOPs and OACs in other systems, given the wide range of biogeochemical composition in inland waters. In particular, there are few measurements of the mass-specific IOPs (SIOPs) in inland waters. Therefore, the focus of this study is to characterize the extent and cause of the spatial variability of the SIOPs within a large turbid freshwater system.
In particular, the main aims of this study are: (1) to improve our quantitative knowledge of the absorption coefficients of dissolved and particulate matter in highly productive and turbid lake systems; (2) to deter-mine the magnitude and sources of spatial variability in the absorption coefficients of the in-water constitu-ents across biogeochemical gradiconstitu-ents. This study builds on previous research on the bio-optical properties of lakes by extending measurements into highly minerogenic waters with marked variability in both phyto-plankton biomass and terrestrial inputs of CDOM. It is anticipated that this work will progress our under-standing of the factors influencing the underwater light field and water-leaving radiative signals in lakes and inform the parameterization and selection of remote sensing algorithms for the retrieval of biogeo-chemical parameters in different lake optical types, including an understanding of the uncertainties and biases on the resulting products.
2. Methods
2.1. Study Site
Lake Balaton is the largest freshwater lake in central Europe by surface area (596 km2) and one of the most
intensively studied. The lake is very shallow with a mean depth of approximately 3 m and, as such, bottom sediments are frequently resuspended in the water column [Herodek, 1986; Presing et al., 2001; Tyler et al., 2006]. This results in high concentrations of suspended mineral particles that can readily exceed 50 mg L21
from resuspension during strong wind events. The lake is composed of four basins and an adjoining a wet-land system (Kis-Balaton) to the west. The main inflow into the lake is the Zala River on the western shore, and the only outflow is a highly regulated channel at Siofok in the east. Nutrient inputs from the Zala River typically produce a pronounced trophic gradient from west to east. While the hydrobiology and hydroecol-ogy have been comprehensively studied, there is presently no published information on the optical proper-ties of Lake Balaton.
At its worst, Lake Balaton experienced hypereutrophic conditions in its westerly basins and eutrophic condi-tions in eastern basins because of increased nutrient loads in the 1970s [Herodek, 1986]. Blooms of filamen-tous cyanobacteria (Cylindrospermopsis raciborskii (Wołoszynska) Seenayya and Subba Raju) dominated the summer plankton community in the 1980s and 1990s [Presing et al., 1996]. Since then, extensive waste water treatment and diversion schemes, the introduction of Kis-Balaton wetlands, closure of nearby farms in 1987, and reduction of fertilizer use have all substantially reduced nutrient loading to Lake Balaton, resulting in lower phytoplankton biomass and improved water quality [Somlyody et al., 1997].
2.2. Water Sampling
Field measurements in Lake Balaton and Kis-Balaton were conducted at 38 stations over a 1 week period in August 2010 (Figure 1), with the aim of collecting data on the spatial variability in the light absorption budget over a gradient from the highly productive, high phytoplankton biomass waters in western basins and Kis-Balaton to the low chlorophyll, low CDOM waters in the eastern basins. Daily average wind speed was also measured during each sampling occasion at automatic stations in order to investigate variations in the IOPs during resuspension events (Table 1; Central-Transdanubian Water Directorate). At each station, a surface water sample was divided into subsamples for subsequent filtration or preservation. Subsamples for the determination of pigment concentrations and laboratory measurements of particulate absorption were filtered on the boat immediately after sample collection under low vacuum pressure through 25 mm GF/F filter papers (Whatman, nominal pore size 0.7 mm). Depending on the turbidity, between 20 and 70 mL of water was filtered. Filter papers were flash frozen in liquid nitrogen for <12 h and stored in a 2808C freezer until analysis (no more than 6 months). Further subsamples for CDOM and total suspended matter (TSM) were kept cool and in the dark on the boat and processed in the laboratory within 24 h. Finally, one sub-sample was collected for phytoplankton enumeration and preserved in Lugol’s solution immediately after collection.
ously, an AC-9 (WET Labs Inc.) in situ spectrophometer was deployed with and without a 0.2 mm AcroPak filter (Pall Corporation) for separation of dissolved and particulate contributions to absorption. Because Lake Balaton is highly turbid and very well mixed, all measurements were made just below the water surface within the first optical depth. Prior to data collection, the AC-S and AC-9 were flushed and debubbled for 5 min. Five minute casts were subse-quently executed at each station with the data recorded to a DH4 data logger (WET Labs Inc.).
An AC-S and AC-9 (WET Labs Inc.) were deployed to collect absorption (a) and attenuation (c) measure-ments. The AC-S collected hyperspectral spectra over 84 wavelengths, from 401 to 755 nm at4 nm resolu-tion, while the AC-9 collected data at nine wavelengths only (412, 440, 488, 510, 532, 555, 650, 676, and 715 nm). The AC-S was only utilized in Basins 1–4, while AC-9 data were collected in Kis-Balaton as well. The AC-S or AC-9 raw data were corrected for the time lag associated with the flow rate for the instrument, then was merged with the CTD data for temperature, salinity, and pressure. Using the CTD data, the effects of temperature and salinity on pure water absorption and attenuation were removed with wavelength-dependent corrections according to Pegau et al. [1997]. To correct for instrument drift, a pure water calibra-tion was subtracted from both attenuacalibra-tion and absorpcalibra-tion data. The proporcalibra-tional scattering correccalibra-tion of Zaneveld et al. [1994] was applied to absorption data to account for inefficient collection of the scattered light within the AC-S or AC-9 reflecting tube. The proportional scattering correction is used here to be con-sistent with recent studies [Leymarie et al., 2010; Slade et al., 2010; Astoreca et al., 2012]. All AC-S and AC-9 data were also screened for any data out with two standard deviations in order to eliminate any error from bubbles or large particles.
2.4. Chlorophyll-a
Frozen GF/F filter papers were thawed from 2808C in the dark prior to analysis for pigments and particulate absorption. Chlorophyll-a (Chl-a) was measured in triplicate via spectrophotometry (Shimadzu UV-1601) after a hot 90% methanol extraction following Iwamura et al. [1970]. The hot methanol method was used here because it has been previously found to provide the most complete extraction of Chl-a for the phyto-plankton types found in Lake Balaton (M. Presing, personal communication, 2014).
The spectrophotometric method was also validated against samples analyzed using high-performance liq-uid chromatography (HPLC). Pigments were extracted in acetone containing an internal standard (apo-caro-tenoate) after Martinez-Vicente et al. [2010] and separated using a reverse-phase Hypersil 3 mm C8 MOS-2 column on Thermo-separationsVC and AgilentVC instruments with photodiode array detection [Barlow et al.,
1997; Llewellyn et al., 2005]. Pigments were quantified against commercial phytoplankton pigment stand-ards (DHI Lab Products, Denmark). The spectrophotometric Chl-a data showed strong agreement with Chl-a results determined using HPLC methods (R250.987, p < 0.001). Results presented hereafter are based on the Chl-a data from spectrophotometry because the HPLC measurements were not replicated.
2.5. Phycocyanin
Phycocyanin was extracted in a solution of 15 mL 0.05 M phosphate buffer (pH 5 6.8). The solution was then subjected to sonication over ice for 15 s (Ultrasonic Homogenizer 4710 Series with microtip and 50% duty cycle, Cole-Parmer Instrument Co., USA) as in Horvath et al. [2013]. The extracts were clarified by
filtration (Whatman GF/C filter) and the absorption was measured on a spectrophotometer (Shimadzu UV-1601, Shimadzu Corp., Japan). Phycocyanin concentrations were calculated using the equations of Siegel-man and Kycia [1978]. Clear outlying values were discarded so the PC concentrations were the mean of (minimum) two replicates.
2.6. Biomass and Phytoplankton Counts
Phytoplankton species were enumerated with an inverted plankton microscope [Uterm€ohl, 1958]. The wet weight of each species was calculated from cell volumes [Nemeth and V€or€os, 1986]. At least 25 cells (or fila-ments) of each species were measured to determine biomass and at least 400 were counted.
2.7. Total Suspended Matter
TSM was obtained by gravimetric analysis. Water (500–1500 mL) was filtered under low-vacuum pressure (<700 mbar, 250 kpa) through a preashed (furnace at 4508C) and preweighed 47 mm GF/C filter paper (Whatman). Following filtration, filter papers were dried for 24 h in a clean oven at 608C and subsequently weighed to obtain TSM. Filters were then placed in a furnace at 4508C overnight and subsequently weighed to obtain particulate inorganic matter (PIM). Particulate organic matter (POM) was calculated as the differ-ence between TSM and PIM.
2.8. Colored Dissolved Organic Matter Absorption
Water samples were filtered into clean glassware through 0.2 mm nucleopore membrane filters (Whatman) and measured according to Tilstone et al. [2002] within 24 h of collection. Absorption of the filtrate was determined on a spectrophotometer (Shimadzu UV-1601) with a 5 cm quartz glass cuvette over the range of 350–800 nm, using MilliQ water as a reference. The absorption coefficient of CDOM (aCDOM) was
calcu-lated using the following equation:
aCDOMð Þ52:303D kk ð Þ=r (4)
where D(k) is the measured absorption at a given wavelength and r is the cuvette path length in meters. A baseline correction was applied by subtracting the mean value of aCDOM(k) in 5 nm interval around 685 nm
[Babin et al., 2003b]. This wavelength was used because there is negligible aCDOMat 685 nm and the effects
of temperature and salinity on water absorption are small [Pegau et al., 1997]. The spectral slope of the CDOM absorption curve (SCDOM) was calculated over the wavelength range of 400–500 nm using an
expo-nential function fitted by nonlinear regression [Twardowski et al., 2004; Perkins et al., 2009].
2.9. Laboratory Measurement of Particulate Absorption
The absorbance of the material on the filter was measured from 350 to 750 nm according to the ‘‘transmittance-reflectance’’ method of Tassan and Ferrari [1998] using a dual beam spectrophotometer (Lambda 2, PerkinElmer Inc.) retro-fitted with a spectralon coated integrating sphere. Absorption was meas-ured before and after bleaching with a 1% solution of NaClO to obtain total particulate absorption (ap
spec
(k)) and absorption by nonalgal particles (aNAP(k)), respectively. The path length amplification correction of
Tas-san and Ferrari [1998] was applied, and absorption by phytoplankton (aph(k)) was calculated as the
differ-ence between ap spec
(k) and aNAP(k). Chlorophyll-specific absorption coefficients (a*ph(k)) were obtained by
dividing aph(k) by the respective Chl-a concentration. An exponential function was fitted by nonlinear
regression to the aNAP(k) spectra, and the spectral slope of aNAP(k) (SNAP) was obtained. Wavelengths 350–
750 nm were used in fitting the exponential function, disregarding the ranges 400–480 and 620–710 nm to avoid any residual pigment absorption, as in Babin et al. [2003b]. The data from three stations (26, 27, and 30) were discarded for the purposes of SNAPcalculation due to spectral artifacts and extremely low values of
aNAP(440).
3. Results
3.1. Comparison of In Situ and Laboratory-Measured Absorption
Bulk absorption coefficients (a(k), m21) measured in situ were compared with summed laboratory
measure-ments of particulate and CDOM absorption (a(k) 5 ap(k) 1 aCDOM(k)). Linear regressions of laboratory and in
situ total absorption measurements at three wavelengths (440, 555, and 676 nm) are shown in Figure 2. There was some agreement between in situ and laboratory measurements, with roughly the same
distribution but some differences in absolute values. In situ measurements generally correlated well with laboratory measurements, with R2 values of 0.743–0.843 (p < 0.001) for the AC-S and 0.871–0.967 (p < 0.001) for the AC-9. However, there was a particular lack of sensitivity in the AC-9 and, to a lesser extent, the AC-S data to variations in absorption at 555 nm. Additionally, in situ measurements overestimated labo-ratory total absorption at these wavelengths by a factor of 1.1–2.1 (AC-S) and 1.5–2.5 (AC-9), with improved agreement in the red portion of the spectrum (676 nm) (Figure 2). In contrast, a previous study by Leymarie et al. [2010] found measured a(k) 1 aw(k) (total nonwater and water absorption) to be underestimated
when using the proportional scattering correction, with the errors highest in the red portion of the spec-trum. In the present study, it is expected that the overestimate of absorption is due to the scattering errors in the AC-9 and AC-S measurements which were not resolved by the proportional scattering correction [Mckee et al., 2008]. Scattering increases at shorter wavelengths, thus a smaller difference (i.e., better agree-ment) would exist between the in situ and laboratory absorption measurements at longer wavelengths. Fur-thermore, there was a marked discrepancy between the two data sets of in situ absorption coefficients (at 440 nm, absolute error 5 0.041–0.31 m21). In general, the lab data more accurately reflected the spatial vari-ability in absorption and had greater sensitivity, particularly in waters with low absorption. Thus, for the pur-pose of this study, laboratory measured absorption and specific absorption coefficients will be considered only, and in situ results are solely presented in the methods for comparison.
3.2. Variability in Optically Active Constituents
The Chl-a and TSM data collected during the sampling campaign are shown in Figure 3 in relation to the annual cycle for these parameters. In all four basins, Chl-a concentrations peaked in early August immedi-ately prior to the sampling campaign due to the development of a cyanobacterial bloom over much of the lake and remained high during the sampling campaign in three of the four basins. TSM concentrations were more variable over the year, with a marked peak in early summer in all basins and a winter peak in Basin 1 due to wind-driven resuspension. The sampling campaign was undertaken during a period with TSM concentrations mostly slightly lower than the annual mean (22 6 21 mg L21) but these were not atypi-cal for Lake Balaton.
The stations sampled in Lake Balaton and Kis-Balaton during summer 2010 demonstrated significant vari-ability in the concentration of optically active constituents (Table 2). Mean Chl-a concentrations ranged from160 mg m23in Kis-Balaton to10 mg m23in Basin 4, and phytoplankton biomass ranged from 70,000 to 2800 mg m23
, over the trophic gradient from west to east across the system. TSM ranged from40 mg L21
in Kis-Balaton to 13 mg L21in Basin 4, with POM comprising the majority of TSM in Kis-Balaton (67%) and PIM comprising the majority of TSM in Basins 1–4. The greatest contribution of PIM to TSM (81%) was observed in Basin 3. Chl-a was strongly linearly correlated with POM (R250.97, p < 0.001, n 5 38) with mean Chl-a:POM 5 0.00395 6 0.00110. Total phytoplankton biomass was also linearly corre-lated with POM (R250.88, p < 0.001, n 5 38) with mean total biomass:POM 5 0.998 6 0.583. The mean PC:POM ratio for all basins is 0.00321 6 0.00145 (R250.75, p < 0.001, n 5 38), although this linear Figure 2. Comparison of total nonwater absorption (a(k) 5 aph(k) 1 aNAP(k) 1 aCDOM(k)) from in situ (AC-S or AC-9) or laboratory methods at (a) 440, (b) 555, and (c) 676 nm. Axes are on
relationship had more dispersion than that for Chl-a or total biomass with POM. Stations in the east of Basin 3 and west of Basin 4 had markedly higher concentrations of TSM because wind-driven resuspension of bot-tom material was more prevalent in these basins during sampling than elsewhere. Chl-a, PC and ratios of POM and PIM at each station are shown in Figure 4, while the gradients in Chl-a and PC concentrations over distance from the Zala River are presented in Figure 5.
Cyanobacteria biomass was found to correlate strongly with measured PC concentrations (R250.97, p < 0.001, n 5 38), and total biomass showed a strong linear relationship with Chl-a (R250.96, p < 0.001, n 5 38). PC con-centrations were highest in Kis-Balaton (156 mg m23) with decreasing concentrations from Basins 1 to 3 (22– 6 mg m23) and a slight increase in Basin 4 (10 mg m23), which corresponds to the abundance of nitrogen fixing cyanobacteria (Figure 4). In almost all stations, cyanobacteria comprised the majority of the phytoplankton (up
Table 2. Mean Biogeochemical Parameters (Standard Deviation) for Each Basin and Kis-Balatona
Kis-Balaton (n 5 3) Basin 1 (n 5 4) Basin 2 (n 5 8) Basin 3 (n 5 8) Basin 4 (n 5 15) Lake Meanb (n 5 35) Units Chl-a 166.51 (83.15) 32.74c,d,e (5.40) 21.12d,e,f (6.71) 8.24c,f (1.91) 10.80c,f (2.30) 15.08 (8.90) mg m23 PC 156.27 (176.68) 22.33d,e (7.41) 15.62d (4.56) 6.19c,f (2.05) 9.95f (2.67) 11.80 (6.19) mg m23 TSM 40.98 (13.28) 14.41g (5.82) 10.36g (1.78) 12.55g (11.23) 15.37g (6.11) 13.47 (7.01) mg L21 POM 27.53 (8.63) 6.09d,e (1.23) 4.71d,e (1.41) 2.41c,f (0.53) 3.41c,f (0.49) 3.78 (1.43) mg L21 PIM 13.44 (11.54) 8.32g (4.79) 5.65g (1.63) 10.14g (10.96) 11.97g (6.04) 9.69 (6.98) mg L21
Total Biomass 70839 (53637) 7062c,d,e
(1780) 3916d,f
(1376) 1854c,f
(603) 2851f
(821) 3348 (1832) mg m23
Cyano Biomass 55876 (58954) 5756c,d,e
(1810) 3456d,f (1163) 1232c,f (759) 2134f (671) 2644 (1658) mg m23 Cyano Biomass 57 (50) 81g (8) 88g (6) 64g (27) 74g (17) 76 (19) % a
Chl-a, chlorophyll-a measured by spectrophotometry; PC, phycocyanin; TSM, total suspended matter; POM, particulate organic mat-ter; PIM, particulate inorganic matmat-ter; Total Biomass, all phytoplankton biomass; Cyano Biomass, cyanobacteria biomass only. Numerical superscripts designate statistically significant differences between the respective parameter in Basins 1–4 using Tukey’s Honest Signifi-cant Difference method (p < 0.01, adjusted for multiple comparisons)
b
Significantly different to Basin 1.
c
Significantly different to Basin 2.
d
Significantly different to Basin 3.
e
Significantly different to Basin 4.
f
Lake Mean includes the 35 stations in the main basins only (not including Kis-Balaton).
g
Not significantly different from any basin.
Figure 3. (a–d) Annual variation in chlorophyll-a (Chl-a) concentrations and (e–h) total suspended matter (TSM) concentrations for 2010 in Lake Balaton Basins 1–4. Solid dots are mean daily values measured during the August 2010 campaign, with error bars indicating standard deviation. Annual data provided by the routine monitoring at a single station in each basin by the Balaton Limnological Institute.
to 96%) with the most abundant species being the nitrogen-fixing cyanobacterium Cylindrospermopsis racibor-skii, which typically comprised over 50% of the cyanobacterial biomass in the lake. Phytoplankton composition at each station is shown in Figure 4c, indicating the dominance of nitrogen-fixing cyanobacteria, with an increasing presence of cryptophytes, chlorophytes, dinophytes, and heterokontophytes in Basins 3 and 4. Figure 4. Barplots of the ratio of (a) particulate inorganic matter (PIM:TSM) and particulate organic matter (POM:TSM), (b) chlorophyll-a (Chl-a) and phycocyanin (PC) concentrations, and (c) phytoplankton community composition as percentage of total biomass at all stations in order from west to east.
Figure 5. Plots of (a) chlorophyll-a (log[Chl-a (mg m23
)]) and (b) phycocyanin (log[PC (mg m23
)]) as a function of distance from the Zala River during the Lake Balaton sampling campaign.
Unusual stations to note include KB1, 30, 31, and 35. KB1 is nearest to the inflow from the Zala River and this area typically has lower biomass than the rest of Kis-Balaton, and generally has lower cyanobacteria bio-mass as well. In 2010, the inflow was high and this prevented the development of cyanobacterial blooms (M. Presing, personal communication, 2014). Stations 30, 31, and 35 had low percentages of N-fixing cyano-bacteria and larger communities of cryptophytes and dinophytes present, although total cell biomass was relatively low (<2100 mg m23). These three stations were sampled on 24 and 26 August 2010, when high wind speeds caused increased turbulent mixing (Table 1), potentially encouraging the dominance of larger phytoplankton cells (e.g., dinoflagellates) at the surface.
Chl-a, CDOM, and TSM were plotted against each other in order to investigate the relationships between these OACs (Figure 6). A weak but significant linear relationship was found between log(TSM) and log(Chl-a) (R2
50.306, p < 0.001, n 5 38; Figure 6a), similar to that reported for Lake Taihu [Zhang et al., 2010] and European coastal waters [Babin et al., 2003b]. However, the increased scatter around this relationship in Lake Balaton is likely attributed to the large proportion of minerals in the suspended matter. Log(aCDOM(440)) and
log(Chl-a) also covaried linearly (Figure 6b; R250.714, p < 0.001, n 5 38), a relationship that has been reported in a range of coastal waters [Babin et al., 2003b]. In Lake Balaton, this is likely due to the fact that both CDOM and Chl-a decrease with increasing distance from the Zala River. However, this relationship broke down in Kis-Balaton and Basin 1, which are closest to the river inflow. No significant linear relationship was reported between log(aCDOM(440)) and log(TSM) (R250.132, p 5 0.025, n 5 38; Figure 6c). Again, in Lake Balaton, much
of the TSM was comprised of PIM due to resuspension, thus a strong relationship with CDOM was not expected.
As the campaign was conducted over several days, any effect of sampling date on the measured concentra-tions was assessed using a generalized linear model, and the only significant relaconcentra-tionship was observed for PIM (p < 0.001). This was likely due to differences in the wind-driven resuspension of mineral particles from the lake bottom. Thus, most of the variability observed in the biogeochemical constituents can be assumed to be due to local differences in fluvial input, biological productivity and in-lake processing of particulate and dissolved material.
3.3. Variability in the Inherent Optical Properties
In general, the IOPs were variable across the system, with the most distinctly different properties exhibited in the westernmost portion, Kis-Balaton (Table 4). As with biogeochemical parameters, the measured (S)IOPs were tested for the effect of sampling date using a generalized linear model. ap(440) was the only parameter
with a significant relationship with sampling date (p < 0.001). Figure 7 illustrates the gradients of aph(440),
aNAP(440), and aCDOM(440) across the lake over increasing distance from the Zala River. Tables 3–5 summarize
the bulk and specific IOPs and statistics for the four basins and Kis-Balaton, and Table 6 provides a summary of the optical properties of other large lakes from selected previous studies for comparison.
The relative contributions of optically active substances to total absorption for the 38 stations sampled on Lake Balaton are shown in Figure 8 for selected wavelengths. At all wavelengths (440, 555, 620, and 675 nm), aNAP(k) was the smallest contributor, consistently making up less than 35% of the total absorption.
At 440 nm, aCDOMcomprised between 33 and 76% and aphbetween 23 and 62% of absorption, while at
555 nm there was a wider range of composition with no clear trend by basin. Absorption at 620 nm included a higher percentage of absorption by phytoplankton (39–95%), although up to 48% and 32% were attributed to aCDOMand aNAP, respectively. At 675 nm over 70% of absorption was due to aphat all sites,
with less than 10% due to aCDOMand up to 23% due to aNAP, although it was noted that CDOM and NAP
had a greater contribution at 620 nm as compared to 675 nm. Thus, aNAPand aCDOMcan contribute
mark-edly to absorption at the wavelengths where PC and Chl-a absorb strongly (620 and 675 nm).
3.3.1. CDOM Absorption
CDOM absorption at 440 nm in Lake Balaton ranged from 0.093 to 2.93 m21. There was a gradient of aCDOM(k) across the lake (Figure 9), with aCDOM(440) highest in the west (Kis-Balaton, 2.82 m21) where the
Zala River enters and lowest in the east (Basin 4, 0.18 m21). Additionally, the difference in aCDOM(440) was
statistically significant between the four basins (Table 5). The mean value for the spectral slope of CDOM (SCDOM) was 0.018 nm21in Kis-Balaton and 0.020 nm21in the Lake Balaton.
SCDOMwas found to decrease exponentially with increasing aCDOM(440) (p < 0.001, n 5 38; Figure 10).
Figure 6. (a) Total suspended matter (TSM) as a function of chlorophyll-a (Chl-a) and CDOM absorption at 440 nm (aCDOM(440)) as a
excluding Basin 4 varied over a much narrower range (0.0183 6 0.000822). There was no significant differ-ence in SCDOMvalues across the four basins (ANOVA p > 0.01). Stations in Basin 4 had the greatest range in
SCDOM(0.015–0.041 nm21), whereas the SCDOMvalues in the western basins and Kis-Balaton had much less
variation (Table 5). This is likely due to the large influence of the Zala River on CDOM concentrations in the west of Lake Balaton, while complex interactions between localized fluvial inputs and increased contribu-tions from authochthonous sources (phytoplankton decomposition) may play a larger role in the more vari-able CDOM concentrations observed in the east of the lake.
3.3.2. Phytoplankton Absorption
Particulate absorption at 440 (ap(440)) was generally dominated by phytoplankton, with aph(440)
contribut-ing up to 90% of the total particulate absorption (Basin 2). Phytoplankton absorption coefficients at 620 and 675 nm exhibited a decreasing gradient from west to east, with the lowest aphat both wavelengths in
Basin 3. Mean aph(675) ranged from 0.078 m 21
(Basin 3) to 1.55 m21(Kis-Balaton), and aph(620) ranged
from 0.038 m21(Basin 3) to 0.85 m21(Kis-Balaton) (Table 5).
Chl-a concentration showed a strong relationship with aph(440) (R250.93, p < 0.001, n 5 38; Figure 11a),
although a greater amount of scatter was noted around 10 mg m23Chl-a with a slightly steeper slope than that found for oceans [Bricaud et al., 1995]. Chl-a was also related to aph(675), with a coefficient of
determi-nation of 0.95 for a fit by least squares (p < 0.001, n 5 38; Figure 11b). Similarly, phycocyanin concentrations were positively correlated with phytoplankton absorption at 620 nm, but phycocyanin itself only explained 81% of the variability in aph(620) (Figure 11c). However, when Chl-a and PC were summed, 93% of the
vari-ability in aph(620) was explained, which reflects the contribution of Chl-a to phytoplankton absorption at
620 nm (Figure 11d).
Figure 7. Plots of (a) log(aph(440) [m21]), (b) log(aNAP(440) [m21]), and (c) log(aCDOM(440) [m21]) as a function of distance from the Zala River during the Lake Balaton sampling
campaign.
Table 3. Absorption Coefficients, Main Basinsa
IOP Min Median Max Mean SD Units aCDOM(440) 0.093 0.32 1.35 0.39 0.30 m 21 SCDOM(400–500) 0.015 0.018 0.041 0.020 0.0057 nm21 ap(440) 0.11 0.31 1.04 0.39 0.24 m 21 ap(675) 0.050 0.12 0.49 0.17 0.11 m21 aNAP(440) 0.00 0.054 0.22 0.059 0.046 m 21 SNAPb 0.011 0.013 0.025 0.015 0.0039 nm21 aph(350) 0.057 0.15 0.46 0.18 0.094 m 21 aph(440) 0.11 0.24 0.91 0.33 0.22 m21 aph(620) 0.020 0.049 0.24 0.073 0.053 m 21 aph(675) 0.048 0.11 0.45 0.16 0.11 m21 a*ph(350) 0.0036 0.012 0.029 0.013 0.0061 m 2 mg21 a*ph(440) 0.010 0.022 0.032 0.022 0.0046 m2mg21 a*ph(620) 0.0029 0.0045 0.0083 0.0047 0.0011 m 2 mg21 a*ph(675) 0.0055 0.0095 0.010 0.010 0.0020 m2mg21 a
Data include four basins of Lake Balaton only (n 5 35).
b
Phytoplankton absorption coefficients were also strongly related to phytoplankton biomass. Using nonlin-ear regression by least squares fit, 88% of the variability in aph(675) was explained by total phytoplankton
biomass, while 73% of that in aph(620) was explained by cyanobacteria biomass. However, using a sum of
cyanobacteria and cryptophyte biomass, a higher percentage of variability in aph(620) was explained (84%).
Higher aph(675) and aph(620) values corresponded with increased total phytoplankton and cyanobacteria
biomass, respectively (Figure 12).
Mass-specific phytoplankton absorption spectra are shown for the four basins and Kis-Balaton in Figure 13. All stations show the distinctive Chl-a absorption maxima peaks at440 nm and 675 nm, and all stations (except for KB1) demonstrate a smaller absorption peak at620 nm due to the presence of phycocyanin. Stations in Basins 3 and 4 also had distinct peaks in the UV portion of the spectrum at approximately 360 nm, which were not visible in the spectra for Kis-Balaton and Basins 1 and 2.
The mean Chl-a-specific absorption coefficient at 440 nm (a*ph(440)) was 0.022 m 2
mg21in Lake Balaton and 0.017 m2mg21in Kis-Balaton (Tables 3 and 4). There was greater variation in mean a*ph(440) across the
basins and Kis-Balaton than observed for mean a*ph(675), ranging from 0.017 to 0.023 m 2
mg21(Tables 4 and 5). The mean a*ph(675) for the four lake basins was 0.010 m
2
mg21(Table 3) and was slightly lower in Kis-Balaton (0.0088 m2mg21; Table 4). Mean values for a*ph(675) showed little variation across the basins
and Kis-Balaton, with a narrow range of 0.0088–0.011 m2mg21(Tables 4 and 5).
Table 5. Mean Absorption Coefficients (Standard Deviation) by Basina
IOP Basin 1 (n 5 4) Basin 2 (n 5 8) Basin 3 (n 5 8) Basin 4 (n 5 15) Units aCDOM(440) 1.09 b,d (0.21) 0.48d,e (0.077) 0.34d,e (0.038) 0.18b,e (0.058) m21 SCDOM(400–500) 0.018f (0.00058) 0.018f (0.00071) 0.019f (0.00076) 0.023f (0.0079) nm21 ap(440) 0.85 b,d (0.16) 0.52c,e (0.24) 0.23b,e (0.074) 0.29b,e (0.064) m21 aNAP(440) 0.13b,d (0.062) 0.043e (0.031) 0.037e (0.050) 0.059e (0.026) m21 SNAP 0.013 f (0.00090) 0.017f (0.0048) 0.017f (0.0048) 0.014f (0.0029) nm21 aph(350) 0.29f (0.048) 0.21f (0.13) 0.14f (0.049) 0.15f (0.073) m21 aph(440) 0.71 c,d (0.13) 0.47c,d (0.24) 0.19b,e (0.050) 0.23b,e (0.058) m21 aph(620) 0.17c,d (0.019) 0.11c,d (0.062) 0.038b,e (0.012) 0.049b,e (0.015) m21 aph(675) 0.36 b,d (0.078) 0.22c,e (0.11) 0.078b,e (0.020) 0.11b,e (0.026) m21 a*ph(350) 0.0092f (0.0020) 0.0096f (0.0039) 0.016f (0.0051) 0.014f (0.0067) m2mg21 a*ph(440) 0.022 f (0.0060) 0.022f (0.0041) 0.023f (0.0048) 0.022f (0.0015) m2 mg21 a*ph(620) 0.0052f (0.00047) 0.0048f (0.0016) 0.0046f (0.00083) 0.0045f (0.0012) m2mg21 a*ph(675) 0.011 f (0.00067) 0.010f (0.0028) 0.0094f (0.0011) 0.010f (0.0021) m2 mg21 a
Numerical superscripts designate statistically significant differences between the respective parameter in Basins 1–4 using Tukey’s Honest Significant Difference method (p < 0.01, adjusted for multiple comparisons).
b
Significantly different to Basin 1.
c
Significantly different to Basin 2.
d
Significantly different to Basin 3.
e
Significantly different to Basin 4.
f
The specific absorption coefficient of phytoplankton showed variability across pigment concentrations (Figure 14). a*ph(440) and a*ph(675) varied by200% and 150%, respectively, within the same basin on
the same sampling date (i.e., Basin 2; Figures 14a and 14b), while a*ph(620) showed less variability and a
slight positive trend over increasing phycocyanin concentrations (Figure 14c). Previous studies in ocean waters have found a*ph(k) to decrease across increasing Chl-a concentrations, due to variations in pigment
composition and pigment packaging [Bricaud et al., 1995; Bricaud, 2004]. Although similar patterns were evi-dent in Lake Balaton at 440 and 675 nm (Figures 14a and 14b), the relationships were not significant.
3.3.3. NAP Absorption
Figure 15 shows the absorption spectra for nonalgal particles for each basin and Kis-Balaton. aNAP(440) was
2–3 times higher in Kis-Balaton than in the lake basins, and was also significantly different between Basin 1 and Basins 2, 3, and 4 (Table 5). In general, marked variability was reported across the lake, with aNAP(350)
ranging from up to 1.6 m21in Kis-Balaton to <0.1 m21 in Basins 3 and 4 (Figure 15). Note that a small amount of residual pigment absorption can occasionally be observed in the spectra in the region of Chl-a absorption (675 nm), due to incomplete bleaching. Residual pigment absorption in the aNAP spectra
propagated to an error of up to 5% in the calculated values of aph(675), therefore this effect on aph(k) was
considered minimal.
aNAP(k) spectra followed a decreasing exponential shape, with a mean slope (SNAP) of 0.0146 nm21
(coeffi-cient of variation 5 26%), ranging from 0.011 to 0.025 nm21across all basins (Table 3). There were no signif-icant differences in SNAPfound between the basins (Table 5). SNAPgenerally declined with an increasing
ratio of inorganic particulates, although the linear relationship was not significant (R250.0507, p > 0.1, n 5 35), with the greatest variability in SNAPat lower ratios of PIM:POM (Figure 16a). SNAPwas negatively
cor-related to aNAP(440) for aNAP(440) <0.1 m21(Figure 16b). At TSM greater than10 mg L21and aNAP(440)
greater than0.1 m21
, SNAPremained relatively constant (0.01 nm21).
Scatterplots of aNAP(440) as a function of TSM and PIM are shown in Figure 17. When applying a linear
regression with a null intercept to aNAP(440) as a function of TSM, a significantly lower slope (0.0069) exists
Table 6. IOP and SIOPs From Selected Previous Studies for Comparison
(S)IOP Min Max Mean SD Units Location Source aCDOM(440) 0.33 45.89 7.99 7.93 m21 Chagan Lake Wang et al. [2011]
aCDOM(440) 0.11 2.00 m21 Western Lake Erie O’Donnell et al. [2010]
aCDOM(440) 0.073 0.234 0.145 0.049 m21 Lake Superior Effler et al. [2010]
0.105 1.607 0.186 0.439
aCDOM(440) 0.27–0.46 1.52– 2.36 0.71–0.91 0.20–0.35 m21 Lake Taihu Zhang et al. [2010]
aCDOM(440) 0.08 0.75 0.23 m21 Lake Erie Binding et al. [2008]
aCDOM(440) 0.27–0.38 1.52–2.36 0.71–0.98 0.26–0.22 m21 Lake Taihu Zhang et al. [2007]
aCDOM(442) 0.43 14.5 2.65 m21 15 boreal lakes Yl€ostalo et al. [2014]
SCDOM(400–500) 0.0165 nm21 Lake Erie O’Donnell et al. [2010]
SCDOM(400–500) 0.0090–0.0111 0.0139–0.0169 0.0107–0.0134 0.0016–0.002 nm21 Lake Superior Effler et al. [2010]
SCDOM(350–500) 0.011 0.025 0.0176 0.0020 nm21 European coastal waters Babin et al. [2003b]
SCDOM(400–500) 0.0178 0.0190 0.0186 nm21 Oneida Lake Effler et al. [2012]
SCDOM(350–700) 0.0155 0.020 0.0182 nm21 15 boreal lakes Yl€ostalo et al. [2014]
SNAP(482–618,712–750) 0.0113 0.0145 0.0128 nm21 Oneida Lake Effler et al. [2012]
SNAP(380–400,480–620,710–730) 0.0089 0.0178 0.0123 0.0013 nm21 European coastal waters Babin et al. [2003b]
SNAP(482–618,712–730) 0.013 nm21 Western Lake Erie Peng and Effler [2013]
a*ph(440) 0.005 0.084 0.018–0.056 0.007–0.021 m2mg21 Three small reservoirs Matthews and Bernard [2013]
a*ph(440) 0.026 0.008 m2mg21 Lake Kasumigaura Yoshimura et al. [2012]
a*ph(440) 0.013 0.505 0.086 m2mg21 Lake Erie Binding et al. [2008]
a*ph(440) 0.033 m2mg21 Laurentian Great Lakes Perkins et al. [2013]
a*ph(440) 0.035 m2mg21 Onondaga Lake Perkins et al. [2014]
a*ph(440) 0.048–0.083 0.012–0.021 m2mg21 Lake Taihu Huang et al. [2015]
a*ph(443)a 0.008 0.095 m2mg21 European coastal waters Babin et al. [2003b]
a*ph(675) 0.0199–0.0274 m2mg21 Lake Chascomus Luis Perez et al. [2011]
a*ph(675) 0.0288 m2mg21 Lake Taihu Sun et al. [2010]
a*ph(675) 0.018 0.005 m2mg21 Lake Kasumigaura Yoshimura et al. [2012]
a*ph(676) 0.002 0.042 0.009 m2mg21 Long Island Sound Aurin et al. [2010]
a*ph(676) 0.008 0.020 0.014 m2mg21 15 boreal lakes Yl€ostalo et al. [2014]
a*ph(670) 0.007 0.157 0.040 m2mg21 Lake Erie Binding et al. [2008]
a*ph(676) 0.0171 m2mg21 Onondaga Lake Perkins et al. [2014]
a*ph(676)a 0.004 0.035 m2mg21 European coastal waters Babin et al. [2003b] a
than that previously found in coastal waters (0.031) by Babin et al. [2003b] (Figure 17a) and thus presumably lower aNAP*(440). The proportion of aNAP(440) to particulate absorption (ap(440)) is additionally correlated
with PIM (Figure 17b), indicating the strong influence by the mineral component of TSM toward nonalgal particle absorption.
4. Discussion
Using the ternary approach proposed by Prieur and Sathyendranath [1981], the relative contributions to the absorption budget were characterized in this study. At 675 nm, at least 70% of absorption is attributed to Figure 8. Ternary plot indicating absorption by phytoplankton (aph), nonalgal particles (aNAP), and colored dissolved organic matter (aCDOM) at (a) 440 nm, (b) 555 nm, (c) 620 nm, and
(d) 675 nm. Unique symbols indicate the basin, and the sampling date 26 August is highlighted in red. Particulate absorption coefficients were measured in the laboratory by a dual beam spectrophotometer, and CDOM absorption was measured by spectrophotometry.
phytoplankton, with up to 30% of absorption accounted for by NAP and CDOM (Figure 8). However, relative contributions were more variable in the blue portion of the spectrum (440 nm), where phytoplankton com-prised between 20 and 70%, CDOM between 30 and 80% and NAP up to 30% of the absorption budget (Figure 8). In contrast, in coastal waters, it has been reported that NAP can form an even greater percentage of up to 80% of the nonwater absorption at 442 nm, although a similar contribution was observed from CDOM [Babin et al., 2003b; Tilstone et al., 2012]. In ocean waters, an approximately equal contribution to nonwater absorption has been measured from CDOM (40–50%) and phytoplankton (30–60%) at 440 nm [Bricaud et al., 2010], although varying over a much narrower range than found in Lake Balaton. This Figure 9. Spectra of absorption by color dissolved organic matter (aCDOM(k)) for all stations in Kis-Balaton and Lake Balaton by basin. Note the different y axis scale for Kis-Balaton.
indicates greater variations in the contributions to nonwater absorption than that reported for ocean waters, suggesting that inland waters may indeed exhibit more variability in optical properties than oceans. It is also important to note that the contribution of NAP to nonwater absorption in Lake Balaton is higher than that reported in ocean waters (e.g., up to 20% and typically below 10% at 440 nm in the South Pacific) [Bricaud et al., 2010]. Contributions of CDOM and NAP were also relatively high at 620 nm (up to 48% and 32%, respectively) and 675 nm (up to 10% and 23%, respectively), wavelengths of particular interest for remote sensing retrievals of PC and Chl-a pigments. Similar instances were reported in some European coastal waters, where up to 60% of total absorption at the PC (620 nm) and Chl-a (665 nm) absorption peaks was due to particulate detritus, and aCDOM occasionally contributed over 80% of absorption at
620 nm [Babin et al., 2003b]. Similar findings were also reported in three South African reservoirs, where up to 60% and 30% of absorption was attributed to CDOM, while NAP contributed up to >90% and 60% at 620 and 675 nm, respectively [Matthews and Bernard, 2013]. The high contribution of NAP and CDOM at these wavelengths therefore must be considered in bio-optical models for pigment retrieval at these wavelengths.
A distinct gradient in optical properties was also observed along the trophic gradient of Lake Balaton. Basin 1 is phytoplankton-dominated water, while Basin 4 is mineral-dominated water, as shown by a decrease in the organic fraction of TSM from west to east, paralleled with a decrease in total phytoplankton biomass (Table 2). Total nonwater absorption (ap(k) and aCDOM(k)) generally decreases from Basin 1 to 4 as the water
progresses from phytoplankton-dominated to mineral-dominated. The significance of this is that Figure 11. Scatterplots of the phytoplankton absorption coefficient (aph) at (a) 440 nm as a function of chlorophyll-a (Chl-a), (b) 675 nm as a function of Chl-a, (c) 620 nm as a function of
phycocyanin (PC), and (d) 620 nm as a function of the summed pigments, PC 1 Chl-a. Chl-a results are by spectrophotometry, and PC results are a selected average of the results by spectrophotometry. Absorption coefficients were measured in the lab by spectrophotometry. Note axes are on logarithmic scale. Solid line is a regression curve by least squares fit, and the dashed line in Figure 11a is the fit from ocean waters in Bricaud et al. [1995].
phytoplankton particles have different absorption properties than mineral particles. Phytoplankton pigments absorb strongly in the blue and red portion of the spectrum [Mobley, 1994], while inorganic particles have the highest absorption in the blue portion of the spectrum and near exponential decrease in absorption across the spectrum [Babin et al., 2003b]. Therefore, this gradient in CDOM, phytoplankton, and mineral particles cre-ates differences in both the quantity and quality of the underwater light field across Lake Balaton.
The IOPs in Lake Balaton generally show marked variability across the basins from the eutrophic western portion where biological particles dominate to the oligotrophic eastern basins with greater relative influ-ence of minerogenic particles. As expected, Kis-Balaton and Lake Balaton had higher CDOM absorption than coastal (aCDOM(443) < 1 m21) [Babin et al., 2003b] or marine waters (aCDOM(375) 5 0.06–4.2 m21,
although values >1 m21are rare) [Bricaud et al., 1981], with levels markedly higher than hyperoligotrophic ocean waters (aCDOM(370) typically <0.04 m21) [Morel et al., 2007]. However, mean CDOM absorption in
Lake Balaton (aCDOM(440) 5 0.58 m21) was comparable with other shallow inland waters, with absorption
coefficients higher than the oligotrophic Lake Superior [Effler et al., 2010] or Lake Erie [Binding et al., 2008], but lower than organic-rich lakes such as Lake Taihu and Chagan Lake in China [Zhang et al., 2007, 2010; Wang et al., 2011].
Comparisons of SCDOMshould be viewed cautiously because there is no commonly agreed standard
wave-length range or method for calculating its value and approaches vary greatly in the literature. However, compared to studies with similar methods of calculation, mean SCDOMfor Lake Balaton (0.020 nm
21
) is gen-erally higher than reported values for marine (0.014 6 0.0032 nm21) [Bricaud et al., 1981], European coastal waters (0.0176 6 0.0020 nm21) [Babin et al., 2003b] and the oligotrophic Lake Superior (0.0107– 0.0134 nm21) [Effler et al., 2010], but comparable to that reported for the shallow and eutrophic Oneida Lake (0.0186 nm21) [Effler et al., 2012]. Fichot and Benner [2012] have shown that SCDOMis a sensitive tracer
of terrigenous dissolved organic carbon (DOC) in river-influenced ocean margins with lower values observed in more terrestrially influenced waters. The pool of DOC in Lake Balaton is likely to be dominated by allochthonous material, certainly in the western parts of the lake closer to the inflow of the Zala River where mean SCDOMwas lower (0.018 nm21).
This study also found aph(675) to correlate linearly with the Chl-a concentration (Figure 11b), although large
variability existed in the aph(k) parameter, especially at stations with higher concentrations of Chl-a.
aph(675) was further related to total phytoplankton biomass (Figure 12a), while aph(620) varied linearly with
the sum of cyanobacteria and cryptophyte biomass (Figures 12b and 12c). aph(620) has not been specifically
investigated in previous studies, and here we show a good relationship with PC concentrations, although a stronger correlation exists at this wavelength with summed PC and Chl-a pigments (Figure 11d). This wave-length (620 nm) is of importance in order to distinguish potentially harmful cyanobacteria blooms, and the correlation shown here for Lake Balaton is evidence for its future application to remote sensing algorithms for phycocyanin retrieval.
Figure 12. Regression of (a) total biomass, (b) cyanobacterial biomass, and (c) cyanobacteria and cryptophyte biomass against the absorption coefficient of phytoplankton (aph) at (a)
675 nm and (b and c) 620 nm, respectively. Absorption coefficients were measured in the laboratory by a dual beam spectrophotometer. Solid lines represent regression curves by least squares fit.
It is important to note that although there was a strong positive dependency of aph(k) on Chl-a (Figures
11a and 11b), this relationship was different from that observed in ocean waters, particularly at the 440 nm peak [Bricaud et al., 1995]. In Lake Balaton, there was greater absorption by phytoplankton at 440 nm per unit Chl-a than found in ocean waters in Bricaud et al. [1995], and this has also been reported in the English Channel [Babin et al., 2003b] and more recently, Lake Onondaga [Perkins et al., 2014] and three South African reservoirs [Matthews and Bernard, 2013]. However, the contrary was found in the North Sea and Western English Channel coastal waters, with slightly lower aph(442) per unit Chl-a [Tilstone
et al., 2012]. Given that the Bricaud et al. [1995] relationship was established over a narrower range of Chl-a (<30 mg m23), it is unsurprising that this relationship is different over the wider range of Chl-a concen-trations found in Lake Balaton (5–250 mg m23). It was suggested by Babin et al. [2003b] that the devi-ance from the Bricaud et al. [1995] relationship was likely a result of differences in phytoplankton cell size, given the widely accepted observation that oligotrophic waters are typically picoplankton-dominated while eutrophic waters are typically microplankton-dominated. In Lake Balaton, the dominant phyto-plankton group (N-fixing cyanobacteria) was composed of mainly Cylindrospermopsis raciborskii, with Aphanizomenon flos-aquae, Aphanizomenon issatschenkoi, Anabaena aphanizomenoides, and Planktothryx Figure 13. Spectra of mass-specific absorption by phytoplankton (a*ph(k)) for all stations in Kis-Balaton and Lake Balaton by basin.
Figure 14. Variability of (a) a*ph(440) and (b) a*ph(675) over concentrations of chlorophyll-a (Chl-a) and (c) a*ph(620) as a function of phycocyanin (PC) concentrations. Specific absorption
agardhii also present. The cell size of these dominant cyanobacteria species was in the region of 100–200 mm, classifying this phytoplankton group as microplankton. Other large species were present in smaller numbers, including dinophytes (e.g., Ceratium hirundinella; 25–100 mm 5 microplankton) and large colo-nial diatoms (e.g., Melosira granulate, 1200 mm 5 microplankton), and the presence of these microplank-ton may account for the greater aph(440) observed in Lake Balaton as compared to ocean waters.
The increased scatter found in the relationship between aph(440) and Chl-a in Lake Balaton at10 mg m 23
Chl-a (Figure 11a) is also evident in coastal waters from the study by Babin et al. [2003b] (see Figure 7f therein), where it appears there is increased scatter in aph(443) from0.3 to 10 mg m
23
Chl-a. In Lake Bala-ton, the greatest variation in aph(440) was observed in Basins 3 and 4, and it is expected that this increased
scatter is a result of variations in the phytoplankton community. Indeed, the eastern basins (Basins 3 and 4) comprised a more diverse phytoplankton community, with generally a greater range of community Figure 15. Spectra of absorption by nonalgal particles (aNAP(k)) for all stations in Kis-Balaton and Lake Balaton by basin. Note the different y axis maximum for Kis-Balaton.
Figure 16. Scatterplots of SNAP(the spectral slope of aNAP) as a function of (a) the ratio of particulate inorganic to organic matter (PIM:POM) and (b) absorption by nonalgal particles at
composition between the stations as compared to the western basins where N-fixing cyanobacteria com-posed anywhere from 14 to 86% of the total biomass (Figure 4). This increased phytoplankton diversity, and thus increased variations in cell size, in the eastern basins would account for the greater variations in aph(440) per unit Chl-a that were observed in this portion of Lake Balaton.
The Chl-a-specific absorption coefficient (a*ph(k)) has been identified as a major source of uncertainty in
accurately retrieving Chl-a in turbid productive waters [Dall’Olmo and Gitelson, 2006]. The mean a*ph(440)
Figure 17. Correlation between (a) total suspended matter (TSM) and absorption by nonalgal particles (aNAP(440)) and (b) particulate
inor-ganic matter (PIM) and the proportion of absorption by nonalgal particles to absorption by particulate matter (aNAP(440):ap(440)). In Figure
17a, dashed line is a linear regression with null intercept indicating the relationship found across the range of TSM in coastal waters in Babin et al. [2003b] (aNAP(440) 5 0.31*TSM), and the solid line is a linear regression with null intercept for Kis-Balaton and Lake Balaton. The
ranged from 0.017 to 0.023 m2mg21across Kis-Balaton and the four lake basins. In comparison, a* ph(443)
varies over a much broader range in ocean (0.01–0.18 m2
mg21; from Figure 1 in Bricaud et al. [1995]) and coastal waters (0.008–0.10 m2
mg21; from Babin et al. [2003b, Figure 6]). However, measured a*ph(440) in
Lake Balaton is within the range measured in three small oligotrophic to hypereutrophic reservoirs (0.005– 0.084 m2mg21) [Matthews and Bernard, 2013]. Comparable a*ph(440) were also reported in Lake
Kasumi-gaura (0.026 m2 mg21) [Yoshimura et al., 2012], while higher coefficients were reported in Lake Erie (0.086 m2 mg21) [Binding et al., 2008] and Onondaga Lake (0.035 m2 mg21) [Perkins et al., 2014]. Mean
a*ph(675) values varied over a narrow range across the four basins and Kis-Balaton (0.0088–0.011 m2mg21).
The mean a*ph(675) in Balaton is on the lower end of the range reported for ocean waters (0.005–0.06 m2
mg21) [Bricaud et al., 1995], and most similar to the coastal waters in the Baltic and Adriatic Seas [Babin et al., 2003b] or Long Island Sound (median a*ph(676) 5 0.010 m2mg21) [Aurin et al., 2010]. While a*ph(675)
in Lake Balaton falls on the low end of the range for highly turbid lakes such as the hypereutrophic Lake Chascomus, Argentina (a*ph(675) 5 0.0199–0.0274 m
2
mg21) [Luis Perez et al., 2011] and eutrophic Lake Taihu, China (mean a*ph(675) 5 0.0288 m2mg21) [Sun et al., 2010], it was similar to the alkaline
hypereutro-phic to mesotrohypereutro-phic conditions in Onondaga Lake, New York, USA (mean a*ph(676) 5 0.0171 m2mg21)
[Per-kins et al., 2014].
Previous studies document that a*ph(k) decreases from oligotrophic to eutrophic waters, due to the ‘‘pigment
package effect’’ and changes in species composition and thus pigmentation [Bricaud et al., 1995]. However, in Lake Balaton, there was no clear trend of decreasing a*ph(440) or a*ph(675) across increasing concentrations
of Chl-a (Figures 14a and 14b), although any trend in a*ph(k) may be unclear in this study simply due to the
relatively small sample size of 38 stations or the relatively small Chl-a gradient in Lake Balaton. Similarly, a*ph(620) showed a narrow range across the basins (0.002–0.008 m2mg21), and a general trend of
increas-ing a*ph(620) across increasing phycocyanin concentrations (Figure 14b). It has recently been suggested that
a*ph(k) varies greatly with phytoplankton species composition; for example, in Lake Taihu a*ph(k) increased
with the succession from chlorophytes to cyanophytes [Zhang et al., 2012]. Lake Balaton was dominated by cyanobacteria during the sampling period, which is possibly why no significant changes in a*ph(k) were
observed between basins. However, there were variations in phytoplankton community composition within the nondominant functional groups, including a greater percentage of chlorophytes, dinophytes, and diatoms (heterokontophytes) in Basins 3 and 4 (Figure 4). In particular, the slightly greater abundance of microplankton in Basin 3 (96%, compared to 87–95% in Basins 1, 2, and 4), including dinoflagellates (Gymnodinium sp., 25 mm), diatoms (Synedra acus v. rad, 110 mm) and chlorophytes (Schroederia robusta, 80 mm; Staurastrum paradoxum, 40 mm), may explain the low mean a*ph(675) measured in this basin (0.0094 m2mg21). Larger
cells are subject to a greater package effect and thus decreased absorption efficiency. Thus, the observed var-iations in a*ph(k) may be a result of changes in pigment packaging within the different cell types due to
varia-tions in cell size with the change in phytoplankton community composition.
The phytoplankton absorption peak in the UV in Basins 3 and 4 corresponds with decreased CDOM absorp-tion in these eastern basins (mean aCDOM(440) 0.34 and 0.18 m21, respectively), compared with the CDOM
absorption in the western basins (mean aCDOM(440) ranges from 0.48 to 2.82 m21). As CDOM absorbs
strongly in the lower wavelengths (see Figure 9), it may serve as a UV protectant for phytoplankton and other organisms. It is possible that the cyanobacteria in the eastern basins of Lake Balaton are compensat-ing for this decrease in CDOM by produccompensat-ing a pigment to absorb harmful UV rays. In a recent study on the Florida Keys, phytoplankton were found to produce mycosporine-like amino acids (MAAs) to absorb ultra-violet (UV) light, with a peak in aphat315–360 nm, to compensate for low CDOM absorption [Ayoub et al.,
2012]. There are inchoative results of MAA production by Cylindrospermopsis raciborskii in Lake Balaton (A. W. Kovacs, personal communication, 2015), and many marine cyanobacteria species have been docu-mented to produce MAAs [Sinha et al., 2007]. An earlier study also details the presence of UV-screening compounds in terrestrial cyanobacteria mats, including MAAs and scytonemin [Cockell and Knowland, 1999]. In freshwater lakes, the literature is scarce, with Microcystis aeruginosa as the only documented cya-nobacteria species found to produce MAAs [Liu et al., 2004]. It is therefore possible that one of the most dominant cyanobacteria in Lake Balaton are producing MAAs or similar photoprotective pigments in response to UV stress in the eastern basins where there are lower concentrations of CDOM.
The SIOPs varied greatly from west to east across Lake Balaton, principally based on the distance from the main source of nutrients and organic matter, the Zala River. Each station was assigned a distance from the
Zala River, and significant differences (p < 0.05) were found where none were previously observed using the four basin designations (Figure 18). a*ph(350) was found to increase with increasing distance from the
Zala River over a decreasing CDOM gradient (Figures 18a and 18c). Given that a single species comprised the dominant phytoplankton over the study period, it is theorized that the change in a*phin the UV portion
of the spectrum is linked to the variable production of photoprotective pigments. Additionally, a significant change in SCDOMwas measured with increasing distance from the Zala River (Figure 18b). The western
basins have a larger terrestrial input to CDOM, given the proximity to the river, and this is reflected in the lower values of SCDOM. It is important to note that this positive correlation is driven by a small number of
sta-tions >60 km from the Zala River, and a subsequent in-depth spatial investigation of SCDOMwould clarify
this relationship. In bio-optical models for retrieval of IOPs and pigments from remote sensing, the aCDOM(k)
absorption spectrum is often derived from an assumed or estimated SCDOM[Lee et al., 2002; Mishra et al.,
2013; Li et al., 2013, 2015]. As such, SCDOMmay be a prime source of error in the parameterization of
analyti-cal models, with further propagation of errors to retrieved pigment concentrations.
Lake Balaton is often characterized by high and heterogeneous concentrations of suspended minerals due to the wind-induced resuspension of dolomite limestone bottom sediments [Tyler et al., 2006], with lake mean PIM concentrations comprising over 70% of TSM. Most of the variability in TSM concentrations in this study was explained by PIM (57%), and significantly more so if the three Kis-Balaton stations are excluded (96%) (Kis-Balaton is dominated by phytoplankton, with POM comprising up to 92% of the TSM at station KB2). The proportion of aNAP(k) was also significantly correlated with PIM, demonstrating the large
contribu-tion from inorganic matter to nonalgal particulate absorpcontribu-tion (Figure 17). A study on mineral absorpcontribu-tion found that mineral absorption in the Irish Sea decreases from blue to red, with a slight increase between 450 and 550 nm [Bowers and Binding, 2006]. This does not seem to be the case in Lake Balaton, and in fact, the spectra sometimes show a slight dip in this wavelength range. This can largely be explained by the dif-ference in sediment types, specifically the siliceous sediments of the Irish Sea [Bryant et al., 1996] contrast-ing with the dolomitic sediments of Lake Balaton [Tyler et al., 2006].
The slope of the aNAP(k) curve (SNAP) was similar across all basins, with a mean value of 0.015 6 0.004 nm21.
Other studies have reported similarly narrow ranges of SNAP, however mean SNAPin Lake Balaton was
dis-tinctly higher than that reported in ocean (0.0094 6 0.0018 [Bricaud et al., 2010] and 0.011 6 0.0025 [Bricaud et al., 1998]) or coastal waters [Babin et al., 2003b; Bowers and Binding, 2006]. Comparable values were reported for the turbid waters of western Lake Erie [Peng and Effler, 2013], although maximum SNAPvalues
in Lake Balaton were higher. Babin et al. [2003b] hypothesized that the observed variations in SNAPin coastal
waters were a result of the differences in the proportion of mineral versus organic matter. Differences related to NAP composition were found in this study, with lower mean SNAPreported in Kis-Balaton where
the highest proportion of organic matter was measured (46–92% POM), while higher mean values were reported in Basins 2 and 3 where PIM comprised up to 90% of TSM (Basin 3). SNAPalso generally declined
with an increasing ratio of inorganic particulates, but followed a distinctly linear decreasing pattern with Figure 18. Variation of (a) log (a*ph(350)) (m2mg21) and (b) log (SCDOM) (nm21) over increasing distance from the Zala River inflow, and (c) relationship between log(a*ph(350)) (m2
mg21) and log (aCDOM(350)) (m 21
